Environ Monit Assess (2012) 184:893–902
DOI 10.1007/s10661-011-2007-9
Spatio-temporal prediction of atmospheric
benzene (Part I)
Tânia Fontes · Nelson Barros
Received: 20 September 2010 / Accepted: 14 March 2011 / Published online: 12 April 2011
© Springer Science+Business Media B.V. 2011
Abstract Benzene is a carcinogenic and genotoxic
pollutant which mainly affects the people health
through the inhalation. Nevertheless, this pollu-
tant is not frequently measured by air-quality net-
works. To solve this problem, some models have
been published to estimate benzene concentra-
tions in the atmosphere. However, the lack of
measures makes difficult the application of com-
plex models in order to get a detailed spatio-
temporal analysis, namely in urban areas. In
this work was developed a simple semi-empirical
model to predict benzene concentrations based
on the ratio of benzene and carbon monoxide
concentrations in order to predict the concentra-
tions of this pollutant in large areas and peri-
ods with lack of benzene measurements but with
higher impact in the human health. The model
was applied to an urban area, the Metropolitan
Area of Oporto, for a period of 12 years (1995–
2006). Monthly correlations between benzene and
carbon monoxide concentrations at Custóias air-
quality station are significant ( p = 0.01) and
higher in winter (rs > 0.7) than in summer (0.3 >
rs > 0.7). Estimate of the monthly ratio of the con-
T. Fontes (B) · N. Barros
Global Change, Energy, Environment
and Bioengineering Center (CIAGEB),
University Fernando Pessoa, Praça 9 de Abril,
349, 4249-004 Porto, Portugal
e-mail: trfontes@yahoo.com
centration of these two pollutants range between
199 and 305. The methodology validation shows
good results (rs = 0.81) which allow, assuming the
availability of carbon monoxide data, the use of
this tool for areas with low benzene recorded data.
The application of this methodology in the study
area shows an annual average trend decrease of
benzene concentrations during the study period,
which may be linked to a general trend decrease
of benzene emissions in European urban areas,
including the study domain.
Keywords Benzene · Carbon monoxide ·
Spatio-temporal · Measure · Prediction
Introduction
To develop a successful policy in urban air-quality
management, the emissions sources identification
and monitoring atmospheric pollutant levels are
essential in order to evaluate the pressures, not
only in health but also on ecosystems and heritage.
Benzene (C6 H6 ), as a carcinogenic and a geno-
toxic pollutant, has a potential impact on human
heath. However, this pollutant still does not have
enough spatial representativity on the Portuguese
air-quality network.
Several studies have been tried to estimate the
benzene atmospheric concentration. Some authors
explore the possibility to replace the “complicated”
894
measurement of benzene concentrations by a
known, easier and cheaper relationship with car-
bon monoxide (CO) (Guerra et al. 1995; Palmgren
et al. 2000; Pfeffer et al. 1995). The studies carried
out in Germany (Pfeffer et al. 1995) and Denmark
(Palmgren et al. 2000) show a linear correlation
between daily averages of carbon monoxide and
benzene, and Guerra et al. (1995) show the coin-
cidence of peaks between carbon monoxide and
benzene in Italy. Several studies also analysed
the [CO]/[C6 H6 ] ratio emissions. De Petris et al.
(1992) assessed carbon monoxide and benzene
emission factor on gasoline composition from the
fleet of medium-aged motor vehicles showing a
[CO]/[C6 H6 ] ratio in the range of 80–200. It is
expected that this relationship was very similar to
road traffic variation, considering the atmosphere
stability and physical–chemical removal processes
of these pollutants on air. Guerra et al. (1995)
found a value of 160 for the [CO]/[C6 H6 ] ratio.
Moreover, according to Carr et al. (2002), the
annual benzene concentrations can be provided
by just two traffic characteristics: volume and per-
centage of congestion. Using this idea, Karakitsios
et al. (2006) developed two models to simulate
benzene concentrations in a canyon. The first
model is based on artificial neural network train
that uses benzene concentrations, traffic density
and vehicle type. The second is a semi-empirical
deterministic model that was based on road trans-
port emission rate, track geometrical characteris-
tics and wind speed and direction. According to
this study, although the neural network techniques
are less capable of reproducing the changing traf-
fic characteristics, they represent a promising tech-
nique to model urban emissions.
The main objective of this paper is to pre-
dict the benzene atmospheric concentrations in
areas and/or periods with lack of benzene mea-
surements but with higher impact in the human
health. Despite the simplicity of the methodology
presented, the innovation of this work consists
of the application scenario, one area and one
period of large dimension with lack of benzene
concentrations data. These results are particularly
important in future applications as epidemiolog-
ical studies and/or risk assessment. Thus, for the
present study, because of the domain dimension
(the Metropolitan Area of Oporto (MAO) have
Environ Monit Assess (2012) 184:893–902
about 817 km2 ), the [CO]/[C6H6] ratio was used
to rebuild 12 years of benzene hourly time series
(1995 to 2006). The results were evaluated using
the limits of human health protection, published
by European Union legislation, to be achieved in
2010 (Directive 2008/50/EC).
Overview of benzene atmospheric measured
in the study domain
Between 1995 and 2006, the number of stations of
air-quality network in MAO has gradually grown,
from two stations in 1995 up to 22 stations in
2006. Most of them are traffic stations (64%), and
the remaining stations are background (27%) and
industrial (9%; Fig. 1).
Nevertheless, in MAO, only one station was
classified as industrial suburban (1021—Custóias),
record the benzene levels and only on the last
4 years (2003 to 2006), which represents an im-
portant limitation to develop a spatio-temporal
analysis in this region. Furthermore, to com-
pare these benzene levels with the annual limit
value for the human health protection, Directive
2008/50/EC define a minimum rate of data col-
lection of 90%, indicated as desirable for this
pollutant, and a uniform distribution over time
(Directive 2008/50/EC). In MAO, this requisite
is not fulfilled due to the low efficiency of data
recorded by the air-quality network (57% in 2003,
74% in 2004, 53% in 2005 and 15% in 2006),
which do not allow an accurate analysis. Nev-
ertheless, assuming these values as hypothetical
representatives during the monitored periods, the
annual averages vary from 1.3–2.5 μg m−3 , which
do not exceed the limit value of benzene (5 μg m−3
in 2010). However, an hourly maximum benzene
value of 76 μg m−3 was recorded (Table 1).
To study the concentrations of benzene in the
surrounding highways in the MAO, the benzene
concentrations were also used, recorded under
a research project—the ImpactAir project (Ref.
POCTI/ESP/47806/2002)—in one of the most im-
portant urban highway that cross the city of
Oporto, the VCI (Barros et al. 2005). In the
absence of data on the air-quality network, sev-
eral passive diffusion monitoring campaigns of
benzene in different periods of representative
Environ Monit Assess (2012) 184:893–902
Fig. 1 Map of air-quality stations located in the Metropolitan Area of Oporto between 1995 and 2006
sub-domains at the VCI were made (Fig. 2).
Background concentrations were evaluated in two
points far away from direct atmospheric benzene
emissions placed on top of two different buildings
with a height of about 40–50 m.
Results show that although Prelada has a higher
number of vehicles than the other sub-domains,
due to its open configuration without any nearby
obstacles or buildings, the dispersion and trans-
port condition are promoted, and consequently,
the observed values of benzene concentration
are relatively low. On the other hand, because
of the traffic pressure and the dispersion limita-
895
tion imposed by the presence of buildings by the
road, the concentrations in Boavista and Amial
sub-domains are higher than Prelada sub-domain
(Table 2). Values measured at points located
nearby the VCI can reach levels up to twice the
measured points 100 m away or compared to the
background concentrations in the points located
on top of the Antas Tower and on the Burgos
Building. Furthermore, concentrations are higher
on streets across VCI, such as Amial Street in
Amial sub-domain and the Boavista Avenue in
Boavista sub-domain. In these points, the mea-
sured concentrations are 50% higher to other
896 Environ Monit Assess (2012) 184:893–902

Table 1 Average monthly benzene concentration (μg m−3 ) and efficiency of data collection in Custóias station (Ref. 1021)
between 2003 and 2006
Months Concentration (µ g,m −3 ) Minimum / A verage / Maximum Efficiency (%)
2003 2004 2005 2006 Average 2003 2004 2005 2006
Jan. − 0 / 6.0 / 49 0 / 1.0 / 12 − 3.5 ± 3.5 − 83.9 75.0 −
Fev. − 0 / 8.2 / 26 0 / 2.2 / 15 1 / 2.8 / 32 5.2 ± 4.2 − 30.0 71.9 29.5
Mar. 0 / 2.9 / 13 0 / 1.6 / 9 0 / 0.3 / 4 1 / 2.5 / 16 1.6 ± 1.3 77.6 63.0 35.8 37.5
Apr. 0 / 1.0 / 9 0 / 1.1 / 14 0 / 0.7 / 7 1 / 2.1 / 12 0.9 ± 0.2 94.2 76.7 44.7 26.9
May. 0 / 1.0 / 13 0 / 0.5 / 4 0 / 0.5 / 5 1 / 2.3 / 16 0.7 ± 0.3 94.5 10.9 72.4 23.5
Jun. 0 / 0.8 / 19 0 / 0.8 / 15 0 / 0.7 / 7 1 / 2.6 / 12 0.8 ± 0.1 100.0 100.0 74.2 17.1
Jul. 0 / 1.3 / 17 0 / 1.1 / 50 − 1 / 2.3 / 13 1.2 ± 0.1 90.9 99.9 − 20.3
A ug. 0 / 1.1 / 10 0 / 1.1 / 76 0 / 2.3 / 13 1 / 2.5 / 8 1.5 ± 0.7 29.6 100.0 26.7 21.2
Sep. 0 / 1.8 / 41 0 / 1.6 / 12 0 / 2.1 / 21 1 / 1.7 / 4 1.8 ± 0.3 95.8 95.6 67.4 3.9
O ct. 0 / 2.1 / 9 0 / 1.0 / 12 0 / 1.5 / 27 1 / 5.3 / 9 1.5 ± 0.6 79.8 99.7 99.5 0.5
Nov. 0 / 2.1 / 10 0 / 2.1 / 14 0 / 1.5 /10 − 1.9 ± 0.3 26.9 69.7 70.0 −
D ec. − 0 / 1.7 / 14 − 1/1/1 1.7 ± 0 − 52.4 − 0.1
A verage 0 / 1.5 / 41 0 / 1.6 / 76 0 / 1.3 / 27 1 / 2.5 / 32 1.5 ± 0.2 57.7 73.6 52.8 15.1
Efficiency: > 90%, < 90%

points located along VCI. Furthermore, in Amial The temporal analysis shows that there was
sub-domain, the presence of a refuelling station a higher concentration in winter (2.6 μg m−3 in
contributes to increase the benzene concentration 2006), when compared with the summer average
in the area (Barros et al. 2005). of 1.6 μg m−3 (a decrease of about 84%). This can

Fig. 2 Measurement of benzene by passive diffusion at VCI

Environ Monit Assess (2012) 184:893–902 897

Table 2 Benzene concentrations in VCI (μg m−3 ), by sub-domain and period

Height Period Limit value Benzene concentrations (μg.m−3 ) to
(m) (μg m−3 )a All domains Boavista Prelada Amial Antas
Min/Avg/Max Min/Avg/Max n Min/Avg/Max n Min/Avg/Max n Min/Avg/Max n
3 Apr/2005 5 1.2/2.0/3.3 1.7/2.4/3.3 14 1.5/2.0/3.1 8 1.5/2.3/2.9 19 1.7/1.8/1.9 3
Jan/2006 1.4/2.6/5.5 1.7/2.8/5.5 13 1.8/2.3/2.9 8 1.9/3.0/3.8 20 2.4/2.6/2.8 3
Set/2006 0.9/1.6/3.5 1.3/2.2/3.5 13 0.9/1.3/1.9 8 1.1/1.7/2.7 20 1.4/1.6/1.9 4
50 Apr/2005 1.2/1.6/1.8 1.8/1.8/1.8 2 – – – – 1.2/1.3/1.4 2
Jan/2006 1.4/1.6/1.9 1.4/1.7/1.9 2 – – – – 1.5/1.5/1.5 2
Set/2006 0.9/1.0/1.2 1.2/1.2/1.2 2 – – – – 0.9/1.0/1.0 2
Min/Avg/Max minimum value/average value/maximum value, n total number of samples
a Annual limit value to human health protection to be achieve in 2010 (Directive 2008/50/EC)

be explained by the increase of solar radiation in
summer, causing an improvement of the rate of
benzene oxidation with OH with the consequent
reduction of benzene concentration. Moreover, in
the study domain, the transport conditions and at-
mospheric dispersion are more limited during the
winter, when compared with summer season
(Barros et al. 2005).
The comparison of benzene concentrations to
the different campaigns (period of 20 days) with
the annual limit value for the human protection
health (Directive 2008/50/EC) is not absolutely
correct due the low temporal representation.
However, for this brief analysis, it was assumed
hypothetically that concentrations are approach-
ing an annual average. Thus, a generic analysis
shows that this limit value is not exceeded in VCI
(Table 2).
Methodology
An important limitation for a spatial–temporal
analysis of atmospheric benzene concentration in
the study domain is the lack of data recorded in
the local air-quality network. In fact, only one sta-
tion has the capability to record the atmospheric
benzene concentration. Thus, a new approach was
developed in order to solve this limitation and
rebuild the benzene data series.
The bibliography presents some methodologies
to estimate these benzene data series; however,
most of them are specific for road traffic surround-
ings. One of the purposed methods has a generic
application and is based on a simple relation
between the concentrations of carbon monox-
ide and benzene. Nevertheless, the simultaneous
recording of concentrations of those two pollu-
tants on the MAO air-quality network just oc-
curred in Custóias station between 2003 and 2006
and with only an annual efficiency lower than
50% which invalidates the direct application of
this proposed methodology.
Thus, a previous analysis of monthly corre-
lations between observed carbon monoxide and
benzene for this air-quality station was done. Af-
ter words, [CO]/[C6 H6 ] ratio has been carried out
from Custóias air-quality station in order to get
a benzene estimation tool from carbon monoxide
observed data to the remaining stations of MAO.
Despite of the lack of data, the methodology
validation was possible by using data from pas-
sive diffusion campaigns done before nearby an
urban highway in Oporto City (VCI). Statistical
correlations and a paired sample t test were used
in order to validate the model to the study do-
main. To do this, the carbon monoxide concen-
tration recorded on three air-quality stations of
MAO, Boavista (Ref. 1027), Antas (Ref. 1028)
and Vila Nova da Telha (Ref. 1031), was used.
These estimates were compared with benzene ob-
served by passive diffusion during the seasonal
campaigns done in Boavista, Antas and in two
points placed in “Antas tower” and in “Burgos
Building” located nearby these VCI sub-domains
(Table 2). Although the places used for com-
parison are not the same, the surrounding and
the type of influences are identical. So, for this
898 Environ Monit Assess (2012) 184:893–902

evaluation process, it was expected that the ob- a temporal screen analysis of these pollutants
served and estimated values are similar, mean- was done. This analysis shows, in the winter, a
ing with the same order of magnitude. To assess high correlation between the two pollutants, with
the model quality, the correlation between ob- a 99% confidence interval and a coincidence of
served and estimated benzene concentration was maximum and minimum values. However, in sum-
determined. mer, the correlation is moderate although with the
same level of confidence (Fig. 3).
Thus, based on the screen analysis and to as-
Results and discussion sess the relationship between benzene and carbon
monoxide concentrations with high confidence
In order to verify the correlation between car- level, the monthly correlation coefficient between
bon monoxide and benzene in Custóias station, these two pollutants was calculated. In order to

a 1200 14
b
60
CO

( g m-3)
ntrations

r = 0,907**
C6H6 12
1000

centrations (µg

Benzene concentrrations (µg.m-3)

50
10
cen

800
onc

8 40
((µg.m-3)
de co

600 6
oxid

onc

30
n mono

4
ne co

400
2
zen

20
bon

200
Benz
Carb

0 2
-2 10
04 0:00

04 6:00

4 12:00

4 18:00

04 0:00

04 6:00

2:00

4 18:00

04 0:00

04 6:00

4 12:00

4 18:00

04 0:00
4 12

0
200

200

004

004

200

200

004

004

200

200

004

004

200

0 1000 2000 3000 4000 5000

1//6/2

1//6/2

6/20

6/20

1//7/2

1//7/2

7/20

7/20

1//8/2

1//8/2

8/20

8/20

1//9/2

`
Carbon monoxide concentration (µg.m -3)
1/6

1/6

1/7

1/7

1/8

1/8

Time (hour variation)

Tim
c 1000 10 d 16
CO
ons

m-3)
µg.m
ntratio

14
C6H6 8
entration (µg.m-3)

800
ncenttrattions (µ
cen

12
6
onc

r = 0,585**
600
de co
(µg.m-3)

10
4
oxid

8
400
Benzene conce
con
m no

2
ec
n mo

6
ene
Carbon

200
nze

0 4
Ben

0 2
-2 2
2:00

8:00

2:00

8:00

2:00

8:00

2:00

8:00
0:00

6:00

0:00

6:00

0:00

6:00

0:00

6:00

0:00
0

0
04 0

04 6

4 12

4 18

04 0

04 6

4 12

4 18

04 0

04 6

4 12

4 18

04 0

04 6

4 12

4 18

04 0
200

200

004

004

200

200

004

004

200

200

004

004

200

200

004

004

200

0 400 800 1200 1600

6//3/2

6//3/2

6/3/20

6/3/20

6//4/2

6//4/2

6/4/20

6/4/20

6//5/2

6//5/2

5/20

5/20

6//6/2

6//6/2

6/6/20

6/6/20

6//7/2

Carbon monoxide concentration (µg.m -3)

6/5

6/5

Time (hour variation)

Fig. 3 Temporal data series and spearman correlation between C6 H6 and CO (μg m−3 ) in the air-quality station of Custóias:
(a, b) winter (6–8/01/2004); (c, d) summer (3–5/06/2004)
Environ Monit Assess (2012) 184:893–902 899

Table 3 Spearman correlation and ratio concentrations among carbon monoxide and benzene,between 2003 and 2006, in
Custóias station
Month Spearman correlation (rs ) Ratio [CO]/[C 6 H 6 ]
2003 2004 2005 2006 Average 2003 2004 2005 2006 Average
Jan. − 0.91** 0.74** − 0.83 − 113 442 − 278
Fev. − 0.89** 0.71** 0.71** 0.77 − 96 280 223 200
Mar. 0.71** 0.78** 0.51** 0.59** 0.65 226 236 339 245 261
Apr. 0.63** 0.72** 0.76** 0.26** 0.59 234 240 348 241 266
May. 0.63** 0.50** 0.54** 0.33** 0.50 238 206 235 221 225
Jun. 0.57** 0.59** 0.43** 0.25** 0.46 247 224 263 184 230
Jul. 0.47** 0.51** − 0.19** 0.39 183 195 − 223 200
Aug. 0.64** 0.47** 0.82** 0.48** 0.60 249 191 239 340 255
Sep. 0.75** 0.75** 0.69** 0.57** 0.69 242 236 202 264 236
Oct. 0.74** 0.57** 0.73** − 0.87** 0.73 196 239 251 110 199
Nov. 0.61** 0.87** 0.80** − 0.76 218 228 290 − 245
Dec. − 0.80** − − 0.80 − 305 − − 305
* p = 5% confidence level, ** p = 1% confidence level
< 0.3 0.3 - 0.7 > 0.7
Correlation: Low, Moderate, Strong

analyse the correlation coefficient between these
two pollutants, a Kolmogorov–Smirnov test (N ≥
30) was applied. The result shows no evidence
that the data series follow a normal distribu-
tion; therefore, a Spearman correlation was used
(Table 3). The average correlation found over all
the study period (annual) was 0.64. During the
winter months, the correlation is stronger and pos-
itive (rs > 0.70), and in summer, the correlation is
moderate (0.3 > rs > 0.7), due to the increase of
OH radicals concentrations in this season. For all
monthly analysis, the correlation is significant to a
99% confidence interval (Fig. 3).
Using the available data of Custóias station,
a monthly average of the [CO]/[C6 H6 ] ratio was
found, ranging from 199 and 305. It was assumed
that this ratio will remain constant for the study
domain (Table 3). Although Guerra et al. (1995)
use an annual average of the [CO]/[C6 H6 ] ratio,
in the present work, a monthly average given the
pollutant seasonal variability was used (Guerra
et al. 1995). Nevertheless, an hourly approach
could not be used because of the low efficiency of
benzene records.
The obtained values for the monthly average
of the [CO]/[C6 H6 ] ratio are somewhat higher
than those found by Guerra et al. (1995). In the
last few years, the benzene emissions to the at-
mosphere was reduced, in particular, with regard
to road traffic activity, due to the new standards,
use of catalytic converters, fuels changes and more
efficient engines introduced in the new cars. This
benzene emissions decrease may take to an in-
crease of the [CO]/[C6 H6 ] ratio over the years.
For the present work, the benzene concentration
estimation of previous years should take into ac-
count this variability factor, although the lack of
data prevents a more thorough analysis.
As done before for the screen [CO]/[C6 H6 ]
ratio evaluation, to the validation model, the
Kolmogorov–Smirnov test (N ≥ 30) was applied
to the observed and estimated benzene series. The
result shows that there is no evidence that data
series follow a normal distribution; therefore, a
Spearman correlation was used (Table 4). The
correlation is strong and positive (rs = 0.81). In ad-
dition, it was used a paired samples t test (N = 37).
It was not possible to detect significant differences
to a 5% significance level ( p = 0.252) between
the mean observed and predicted values (Fig. 4).
The estimates are closer to the observed values
in winter than in summer. This behaviour may be
900 Environ Monit Assess (2012) 184:893–902

Table 4 Seasonal average benzene concentration (μg m−3 ), observed and estimated, in January and September of 2006, to
Boavista and Antas sub-domain and background concentration in Oporto City
Period Type Average benzene concentration (μg m−3 )
Boavista Antas Background
10-31/01/2006 Observeda 2.1 ± 0.8 2.5 ± 0.9 1.6 ± 0.6
Estimatedb 2.2 1.8 1.4
12-29/09/2006 Observeda 1.5 ± 0.5 1.7 ± 0.6 1.0 ± 0.4
Estimatedb 2.0 2.3 0.7
a Average concentration, measured with passive diffusion at 150 m distance of VCI, to Boavista and Antas sub-domain and

to background concentrations (50 m height)

b Average estimated concentration to air quality monitoring stations of Boavista (Ref. 1027), Antas (Ref. 1028) and Vila

Nova da Telha (Ref. 1031)

explained by the increase of OH radical and other
oxidants in atmosphere during this season which
promotes the decrease of benzene concentrations.
With the application of the validated methodol-
ogy defined previously, the study of 19 air-quality
stations on the MAO domain throughout the pe-
riod considered was possible (1995 to 2006) which
represents an estimation of 108 benzene time se-
ries. However, only 54% was considered of these
annual series due to carbon monoxide limitations
on data collection related with lack of efficiency
(less than 90% hourly records) and not uniform
data distributions over time according to Directive
2008/50/EC. In the selected time series, only 2% of
the data exceed the annual benzene limit value to
human health protection to be achieved in 2010
(5 μg m−3 ). The estimated benzene concentra-
tions in the traffic stations ranges between 1.7 and
8.9 μg m−3 with an average value of 2.5 μg m−3 .
To background and industrial stations, the av-
erage is 1.8 (1.4–2.2) and 1.5 (1.2–1.7) μg m−3 ,
respectively (Table 5). Another important finding
is that the annual average has been decreasing.
The reasons for this trend can be linked to the
variability of weather conditions, which have an
important influence on the atmospheric disper-
sion capability or in changes on benzene emissions
patterns in urban areas, like the variations of the

Fig. 4 Spearman Estimated concentration

correlation between
4.00
observed and estimated
benzene concentrations
(μg m−3 ) to 95%
confidence level

3.00

2.00

R Sq Linear = 0.691

1.00

1.00 2.00 3.00 4.00

Observed concentration
 Environ Monit Assess (2012) 184:893–902 901

Stations number
vehicles number and/or their characteristics (age,
fuel type, engine technical specifications and pres-
by year ence/absence of catalyst).
Table 5 Monitoring efficiency and annual average estimative of benzene concentration (μg m−3 ) in the air-quality stations of MAO between 1995 and 2006

15
15
2
2
2
3
6
10
11
14
14
14

19
1034 1040 1043

2.3
2.1
2.3

2.2
Conclusions

1.7
1.6
0.7
1.9

1.5
The air-quality network of the MAO has a strong
limitation for benzene observation. In fact, only

2.8
2.0
2.3
1.7
1.9
1.2
2.2

2.0
one station has the capability to record the at-
mospheric benzene concentration. In order to
1029 1030 1031 1032 1033

2.6
2.2
3.7
2.4
2.6

2.7
solve this limitation, a methodology based on the
[CO]/[C6 H6 ] ratio have been implemented.
2.2
2.0
2.5
2.7
2.2
2.1
2.2

2.3
The results also shows that the monthly corre-
lations between observed carbon monoxide and
1.8
1.6
8.8
1.6
1.6
1.4
1.4
2.0

2.5

benzene at Custóias air-quality station are always

significant ( p = 0.01) thought higher in the win-
2.2
1.9
2.6
1.9
1.8
2.3

2.1

ter (rs > 0.7) than in summer (0.3 > rs > 0.7).
After words, a ratio between carbon monoxide
2.4
2.2
2.3

and benzene has been carried out, in order to get

a benzene estimation tool from carbon monox-
1026 1027 1028

1.7
2.8
4.2
2.8

2.9
2.4
2.6
3.0

2.8

ide observed data to the other stations of MAO.

These estimations were carried out monthly using
2.7

4.1
2.9
2.6

2.1
2.2
2.2

2.7

a [CO]/[C6 H6 ] ratio from Custóias air-quality sta-

tion. These monthly average ranges between 199
2.4

2.6
2.2
2.6
2.9

2.5

and 305.
Despite the lack of data, the validation of this
1017 1018 1019 1020 1021 1022 1024 1025

1.3

1.5
1.2
1.3
1.4
1.3
1.3

methodology is possible by using data from pas-

sive diffusion campaigns done before nearby an
2.0
2.8
2.3
1.9
2.0
2.0
1.7
2.1

urban highway in Oporto City (VCI). The ap-

Background

plication of Spearman test to this series shows a

strong and positive correlation (rs = 0.81) and
2.5
4.4
2.1
2.0
2.4
2.6
2.0
2.6

the paired samples t test (N = 37) shows that a

2.0
1.7
1.5
1.6
2.4
1.8
1.6
1.8

5% significance level was not possible to detect

significant differences between observed and pre-
Industrial,

dicted values ( p = 0.252).

4.6
3.8

4.2

The temporal analysis shows that the higher

< 90%

concentration was in winter (2.6 μg m−3 in 2006),

8.9
3.9

6.4
Station reference

when compared with the summer average of

1.6 μg m−3 due to variations with rate of ben-
4.8
1.9
2.0
2.7

2.3
1.6
1.7

2.4

Traffic,

zene oxidation with the OH. Between 1995 and

> 90%,
6.5
6.7
3.6
3.2

4.1
3.2
2.3
1.1

3.8

2003, the annual average estimation of benzene

concentrations shows a consistent decrease. This
Station year type

reduction may be linked, either to the variability

Station type:

of meteorological conditions, which force the dis-

Efficiency:

persal and transport capacity of the atmosphere

Average

and/or changes of benzene emissions patterns.

1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006

Nevertheless, the annual limit value to human

902
health protection of benzene was not exceeded in
none estimated series (5 μg m−3 in 2010).
The methodology presented allowed a spatio-
temporal analysis of benzene concentrations in
the study domain. However, the limitations im-
posed by the representativeness of the data of
observation network may limit the application of
the methodology developed. Thus, other methods
of diagnostic analysis, as air quality modelling, can
be used.
Acknowledgements Thanks to Fundação para a Ciên-
cia e Tecnologia (FCT) for supporting the Ph.D grant
(Ref.◦ SFRH/BD/19027/2004) conceded to Tânia Fontes
and the financial support of ImpactAir Project (Ref.◦
POCTI/ESP/47806/2002).
References
Barros, N., Fontes, T., Brás, C., & Cunha, L. M. (2005).
Population exposure to urban highway traffic emis-
sions. In Environmental health risk III (pp. 221–230).
Bolonha:WitPress.
Environ Monit Assess (2012) 184:893–902
Carr, D., von Ehrenstein, O., Weiland, S., Wagner, C.,
Wellie, O., Nicolai, T., et al. (2002). Modeling annual
benzene, toluene, NO2 , and soot concentrations on
the basis of road traffic characteristics. Environment
Research, 90(2), 111–18.
De Petris, G., Giglio, V., Police, G., & Prati, M. V. (1992).
The inf luence of gasoline formulation on combustion
emissions in spark-ignition engines (pp. 343–53) Mod-
elling and Measurements for S.I. Combustion and
Emissions.
Directive 2008/50/EC. Ambient air quality and cleaner air
for Europe.
Guerra, G., Lemma, A., Lerda, D., Martines, C., Salvi, G.,
& Tamponi, M. (1995). Benzene emissions from motor
vehicle traffic in the urban area of Milan: hypothesis of
health impact assessment. Atm Environmental, 29(23),
3559–69.
Karakitsios, S. P., Papaloukas, C. L., Kassomenos, P. A.,
& Pilidis, G. A. (2006). Assessment and prediction of
benzene concentrations in a street canyon using ar-
tificial neural networks and deterministic models: their
response to “what if” scenarios. Ecological Modelling,
193(3–4), 253–70.
Palmgren, F., Berkowicz, R., & Skov, H. (2000). National
Environmental Research Institute. Ministry of Envi-
ronment and Energy, NERI Techical report n◦ 309.
Pfeffer, H. U., Friesel, J. E. G., Beier, R., & Ellermann,
K. (1995). Air pollution monitoring in street canyons
in North Rhine-Westphalia, Germany. The Science of
the total environment, 169(1–3), 7–15.